Large walls and other typical boundaries strongly influence neural activity related to navigation and the representations of spatial layouts. They are also major aids to reliable navigation behavior in young children and nonhuman animals. Is this because they are physical boundaries (barriers to movement), or because they present certain visual features. such as visually extended 3D surfaces? Here, these 2 factors were dissociated by using immersive virtual reality and real boundaries. Eighty adults recalled target locations in 1 of 4 environments: plywood, where a virtual wall coincided with a large piece of real plywood; pass through, where the virtual wall coincided with empty space and participants could pass through it; pass over, where the virtual wall was projected downward to be visible underneath a transparent floor; and cones. where the walls were replaced with traffic cones. One condition had features that were boundaries and looked like boundaries (plywood); 2 had features that were not boundaries but looked like boundaries (pass over/through); and 1 had features that were not boundaries and did not look like boundaries (cones). The precision and bias of responses changed only as a function of looking like a boundary. This suggests that variations in spatial coding are more closely linked to the visual properties of environmental layouts than to whether they contain physical boundaries (barriers to movement).